When people say ‘I don’t know’ they may get a range on interactional jobs accomplished other than claiming lack of knowledge. For instance, they may resist a line of questioning or stir away from a topic (Beach & Metzger 1997; Hutchby 2002). They may also use the expression with reduced semanticism as an interaction-organizational marker (for an overview see Lindström, Maschler & Pekarek Doehler 2016). In this talk, I examine the use of expressions of the type ‘I don’t know’ as markers for projecting dispreferred responses, i.e., responses that disagree or disalign with the terms set up by the prior speaker’s action.

Based on data from naturally occurring conversations in five languages from distinct language (sub)families (Czech, French, Hebrew, Mandarin, Romanian), I document a multimodal practice that speakers deploy recurrently when providing a dispreferred response. The practice involves the verbal delivery of a turn-initial expression corresponding to ‘I don’t know’ and its variants (‘dunno’) coupled with gaze aversion from the prior speaker; this practice differs from speakers’ ‘literal’ use of ‘I don’t know’, which tends to be delivered with gaze on recipient – at least in ordinary conversation. Through the said practice, respondents preface a dispreferred response, alerting co-participants to incipient resistance to the constraints set out or to the stance conveyed by the prior action. The ‘multimodal assembly’ is found in dispreferred responses to questions, assessments, proposals and informings. This provides one piece of evidence for how participants’ multimodal conduct maps onto one of the basic organizational principles of social interaction: preference organization – and how it does so in a similar manner across different languages.

The findings deepen our knowledge of the type of turn-initial particles pertaining to preference organization, and shed further light on how verbal and bodily conduct interface in social interaction. They amplify prior observations according to which gaze aversion is found with dispreferred responses (Kendrick & Holler 2017, Robinson 2020) by showing that this association is valid across a range of sequence and action types, and across genetically different languages. Such evidence opens a window onto cross-linguistic, cross-modal, and cross-cultural consistencies in human interactional conduct. Yet, it also begs the question ‘why’: Why would an epistemic expression like ‘I don’t know’ lend itself to the purpose of prefacing dispreferred responses? I address this question in the conclusion of this talk.

This presentation has grown out of a collaboration with H. Polak-Yitzhaki, X. Li, I. Stoenica, M. Havlík, and L. Keevallik.